Wrong Msg.3 allocation

at the end nr_get_Msg3alloc(), an allocation for Msg.3 is searched. However, if a UE connects and there is PUSCH transmissions ongoing, the gNB often asserts with.

Assertion (!vrb_map_UL[i + ra->msg3_first_rb + ra->msg3_bwp_start]) failed!
In nr_add_msg3() /oai-ran/openair2/LAYER2/NR_MAC_gNB/gNB_scheduler_RA.c:1082
RB 0 in   18.17 is already taken, cannot allocate Msg3!

The exact root-cause is not clear to me, this should be fixed.

Steps to reproduce:

1. connect a UE in RFsim
2. create traffic
3. connect a second UE

added 5G-NR label

changed title from Inflexible Msg.3 allocation to Wong Msg.3 allocation

mentioned in merge request !1674 (merged)

changed title from Wong Msg.3 allocation to Wrong Msg.3 allocation

This is strange because MSG3 is scheduled before UL traffic (feedback time between DCI and UL is higher for MSG3)

I also confirm that when using 2 COTS UE this Assertion is the main reason for the gNB to not be able to run reliable and for long periods of time.

In rf-simulator it helped to increase the prach_dtx_threshold, so we had less fake PRACH detections. This might probably also help in your case. But of course this is only a workaround to make this assertion less likely...

The issue seems to be the assertion not checking at symbol level. There is no actual overlapping but CSI report in the same slot.

mentioned in merge request !1685 (merged)

@sli this issue is fixed by !1685 (merged)

So, this issue can be closed as soon as Merge Request 1685 is merged.

closed